Open omershlo opened 4 years ago
How performant does one expect the library to be compared to, say Sundersoft, which won the Chia networks competition?
This is a good question! 1) The foundations of this library are built using Pari which is optimised c code for class group operations 2) However, I suspect that our translation from/to Pari to/from our GMP bigint is not so efficient. 3) BigInt operations are suppose to be reasonably efficient.
Overall I would assume the code will be slower than Sundersoft, but : 1) probably within same order of magnitude 2) specifically for VDF, the gap quality between generating and solving will be the similar 3) There is a LOT of room for improvement in this library
This is a good question!
- The foundations of this library are built using Pari which is optimised c code for class group operations
- However, I suspect that our translation from/to Pari to/from our GMP bigint is not so efficient.
- BigInt operations are suppose to be reasonably efficient.
Overall I would assume the code will be slower than Sundersoft, but :
- probably within same order of magnitude
- specifically for VDF, the gap quality between generating and solving will be the similar
- There is a LOT of room for improvement in this library
poanetwork also has this discussion: https://github.com/poanetwork/vdf/issues/13
MPIR may be a better choice over GMP & Pari
add benchmarks for different parameters and measure the absolute and relative times of
Eval
andVerify
If possible, compare it to: https://github.com/harmony-one/vdf https://github.com/poanetwork/vdf